[AMD][AgentX] kimik2.7: explicit AITER-MLA=0 + EP1-vs-EP8 900s tune sweep

- Set VLLM_ROCM_USE_AITER_MLA=0 explicitly (64 KV heads incompatible with the
  ROCm AITER-MLA kernel; was only working via implicit fallback).
- Add kimik2.7-fp4-mi355x-vllm-agentic-tune config key: TP8+LMCache, EP1 vs EP8
  at conc[16,32], 900s. Probes whether expert parallelism improves perf (current
  record sweep runs EP=1 on a large MoE).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: hyukjlee

benchmarks/single_node/agentic/kimik2.7_fp4_mi355x.sh

@@ -53,6 +53,12 @@ if [[ "$version" == "" || ${version:-0} -lt 177 ]]; then
 fi
 
 export VLLM_ROCM_USE_AITER=1
+# amd/Kimi-K2.7-Code-MXFP4 has 64 KV heads; the ROCm AITER-MLA kernel only
+# supports 16 or 128 heads, so it MUST be disabled for this model. Previously
+# left unset (inherited master=1) and only worked via vLLM's implicit backend
+# fallback -- make it explicit so a fallback/backend-selection change (or the
+# --block-size=1 AITER-MLA forcing path) can't turn this into a hard crash.
+export VLLM_ROCM_USE_AITER_MLA=0
 export VLLM_ROCM_QUICK_REDUCE_QUANTIZATION=INT4
 # Avoid intermittent symm_mem all-reduce rendezvous hang at engine init on
 # MI35x nodes (see KIMIK27_CONC64_LMCACHE_RUNBOOK error #4).

configs/amd-master.yaml

@@ -571,6 +571,27 @@ kimik2.7-fp4-mi355x-vllm-agentic:
       - { tp: 8, kv-offloading: dram, kv-offload-backend: { name: mooncake, version: "source-main" }, conc-list: [16, 32, 48] }
 
 
+# Perf-tuning experiment key (same recipe as the record sweep above, via
+# model-prefix kimik2.7). Wave 1: does expert parallelism help? Compares EP1
+# (current) vs EP8 at TP8+LMCache. Dispatched at 900s via e2e-tests duration
+# override to verify the perf delta before committing a full 3600s run.
+kimik2.7-fp4-mi355x-vllm-agentic-tune:
+  image: vllm/vllm-openai-rocm:v0.24.0
+  model: amd/Kimi-K2.7-Code-MXFP4
+  model-prefix: kimik2.7
+  runner: cluster:mi355x-amds
+  precision: fp4
+  framework: vllm
+  multinode: false
+  scenarios:
+    agentic-coding:
+    - dram-utilization: 0.80
+      duration: 900
+      search-space:
+      - { tp: 8, ep: 1, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "aaf7c0d3" }, conc-list: [16, 32] }
+      - { tp: 8, ep: 8, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "aaf7c0d3" }, conc-list: [16, 32] }
+
+
 kimik2.5-fp4-mi355x-atom:
   image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.4_202607091539
   model: amd/Kimi-K2.5-MXFP4